metadata

license: cc-by-nc-sa-4.0
task_categories:
  - image-segmentation
language:
  - en
tags:
  - medical
  - biology
pretty_name: CholecSeg8k
size_categories:
  - 1K<n<10K

Description:

paper | kaggle

The CholecSeg8k dataset, an extension of the Cholec80 collection, includes 8,080 carefully annotated images from laparoscopic cholecystectomy surgeries, selected from 17 video clips in Cholec80. Each image in CholecSeg8K is pixel-level annotated for thirteen different surgical elements. The dataset is efficiently organized in a directory structure, featuring 101 folders, each containing 80 frames at a resolution of 854x480, along with three types of masks for each frame: a color mask for visualization, an annotation tool mask, and a watershed mask for simplified processing. This comprehensive dataset, freely available under the CC BY-NC-SA 4.0 license, is a critical resource for advancing the field of computer-assisted surgical procedures.

Loading the data:

First install the datasets library, then run the following code,

from datasets import load_dataset

dataset = load_dataset("minwoosun/CholecSeg8k", trust_remote_code=True)

Simple demo:

This short demo shows how to load the data and directly visualize an image along with the corresponding masks.

from datasets import load_dataset
import matplotlib.pyplot as plt

dataset = load_dataset("minwoosun/CholecSeg8k", trust_remote_code=True)

def display_image(dataset, image_index):
    '''Display the image and corresponding three masks.'''
    
    fig, axs = plt.subplots(2, 2, figsize=(10, 10))
    
    for ax in axs.flat:
        ax.axis('off')
    
    # Display each image in its respective subplot
    axs[0, 0].imshow(dataset['train'][image_index]['image'])
    axs[0, 1].imshow(dataset['train'][image_index]['color_mask'])
    axs[1, 0].imshow(dataset['train'][image_index]['watershed_mask'])
    axs[1, 1].imshow(dataset['train'][image_index]['annotation_mask'])
    
    # Adjust spacing between images
    plt.subplots_adjust(wspace=0.01, hspace=-0.6)
    
    plt.show()

display_image(dataset, 800) # video index from 0 to 8079

Citation (BibTex):

@misc{hong2020cholecseg8k,
      title={CholecSeg8k: A Semantic Segmentation Dataset for Laparoscopic Cholecystectomy Based on Cholec80}, 
      author={W. -Y. Hong and C. -L. Kao and Y. -H. Kuo and J. -R. Wang and W. -L. Chang and C. -S. Shih},
      year={2020},
      eprint={2012.12453},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

Data card contact:

Min Woo Sun (minwoos@stanford.edu)